Improved algorithm for material characterization by simulated indentation tests

The paper involves the establishment of a neural network model with improved algorithm for reverse analysis of simulated indentation tests considering the effects of friction on the contact surfaces. Extensive finite element analyses covering a wide practical range of materials obeying power law strain-hardening have been carried out to simulate the indentation tests. The results obtained from the simulated dual indentations using conical indenters with different geometries considering effects of friction are adopted in the training and verification of the least squares support vector machines involving structural risk optimization. The characteristics and performances of the neural network model for this class of problems are given and deliberated. The tuned networks are able to predict accurately the mechanical properties of a new set of materials. The approach has great potential for the applications on the characterization of a small volume of materials in micro-and nano-electromechanical systems (MEMS & NEMS).